acdream/tests/AcDream.Core.Tests/Vfx
Erik 845d70248c weather(phase-6a): port retail PhysicsScript runtime
The central runtime for every client-visible scripted effect server
triggers via PlayScript (opcode 0xF754) — spell casts, emote
gestures, combat flinches, AND lightning flashes during storms.
Previously acdream parsed PhysicsScript from the dat (via DRW) but
had no runner; the PlayScript stub in ParticleSystem.cs was a
no-op.

Decompile provenance (`docs/research/2026-04-23-physicsscript.md`,
`docs/research/2026-04-23-lightning-real.md`):

  FUN_0051bed0 — play_script(scriptId) public API — resolves the
                 dat id, queues into the owner's ScriptManager list.
  FUN_0051be40 — ScriptManager::Start — alloc 16-byte node
                 {startTime, script*, next}.
  FUN_0051bf20 — advance one hook, schedule next fire by next
                 hook's StartTime.
  FUN_0051bfb0 — per-frame tick: while head.NextHookAbsTime ≤
                 globalClock, fire via vtable dispatch.

Port choices:
  - Flat List<ActiveScript> vs retail linked list — iteration is
    simpler, N is small.
  - Scripts keyed by (scriptId, entityId) — replay replaces instead
    of stacking, matches retail's "play_script on the same obj
    doesn't double-schedule".
  - Anchor world pos cached at Play() time — good enough for
    short-lived effects (lightning, spell casts). Callers that
    need fresh positions for long emote animations can Play()
    again each frame (idempotent).
  - Constructor takes Func<uint, PhysicsScript?> resolver so tests
    don't need DatCollection; production uses the DatCollection
    overload that wraps Get<PhysicsScript> with null-on-fail.
  - CallPESHook recurses Play() with Pause baked into the
    sub-script's StartTimeAbs. Matches retail semantics where
    nested scripts fire on the NEXT tick (list iteration order).

Diag: ACDREAM_DUMP_PLAYSCRIPT=1 logs every Play() and every fire as
[pes] lines. Use this to identify the actual script IDs your ACE
server is sending so we can confirm the lightning pipeline when the
server sends a strike.

Test coverage (9 new tests, all passing):
  - unknown script returns false, zero id silent-ignore
  - hooks fire in order at their scheduled times
  - entityId + anchor pass through to sink
  - replay same (scriptId, entityId) replaces, doesn't stack
  - different entities run independently
  - StopAllForEntity cancels that entity's scripts only
  - CallPES nested spawn semantics (fires next tick)
  - CallPES with Pause delays correctly

No GameWindow wiring yet — Phase 6b handles the 0xF754 packet
handler and Phase 6c plugs the runner into the frame loop.

Build + 742 tests green.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-04-24 11:20:39 +02:00
..
ParticleSystemTests.cs feat(vfx): Phase E.3 particle system + hook wiring + registry 2026-04-18 16:48:17 +02:00
PhysicsScriptRunnerTests.cs weather(phase-6a): port retail PhysicsScript runtime 2026-04-24 11:20:39 +02:00